Towards Scalable 3D Integration of 2T-nC FeRAM with Hundreds of Layer Stacking

In this work, we study the limits of the number of capacitors and read history dependence in a 2T-nC ferroelectric random-access memory (FeRAM) cell, paving the way for its high-density integration toward hundreds of stacked layers. Through a comprehensive experimental and simulation study on the scaling behavior of the 2T-nC FeRAM architecture, we demonstrate: (i) successful fabrication of 2T-64C cells with robust memory operation and clearly distinguishable '0' and '1' states, even in 64-capacitor configurations; (ii) that the parasitic capacitance of the floating node originates predominantly from the linear component of the ferroelectric capacitor, and its impact on n-scaling - due to degraded sense margin - can be mitigated by floating unselected capacitors with enough TΩ isolation; (iii) that sharing write and read transistors among n capacitors introduces a read history dependence issue due to fluctuating floating node voltage (V<inf>FN</inf>); and (iv) that a proposed FN discharge scheme can effectively eliminate read-sequence dependence, at the cost of reduced read endurance.